[llvm] [CodeGen] Don't allow function alignment less than the target's minimum (PR #90415)

[Eli Friedman via llvm-commits]

efriedma-quic wrote:

> I could add "Fi32" to the AArch64 data layouts. (Any pointers to how to do this most cleanly?)

Should be Fn32. ("i" is only for targets where function pointers aren't simple pointers to code... like Arm and PPC.)

The datalayout is in two places: clang/lib/Basic/Targets/AArch64.cpp, and llvm/lib/Target/AArch64/AArch64TargetMachine.cpp .  Then, ideally, you update UpgradeDataLayoutString in llvm/lib/IR/AutoUpgrade.cpp so auto-upgraded IR inherits the right behavior.

> (newbie Q: Should I open a different PR or repurpose this one?)

In this case, I'd create a new PR because the new approach is so much different.

https://github.com/llvm/llvm-project/pull/90415